Drone, LiDAR & GeoTIFF — 80% smaller, drop-in
Standard files in. Standard files out. Survey, agriculture and GIS data shrinks by 80–99%, with CRS, attributes and geometry preserved.
Drop a .tif, .las, .laz, .geojson or zipped .shp. We compress it in your browser flow. No credit card.
The math at 10 TB / month
A drone-mapping or ortho-mosaic shop ingests roughly 10 TB of new GeoTIFF / LAS per month. Here's what that costs to keep online for a year:
| Vendor | Storage (10 TB) | Notes | Annual cost |
|---|---|---|---|
| Pix4D Cloud Advanced | included up to 1 TB; €0.10/GB over | €599/mo plan + 9 TB overage | ~€18,000/yr |
| Esri ArcGIS Online | credits-based | ~$1,200/TB-yr typical | ~$12,000/yr |
| AWS S3 + GeoServer | $0.023/GB-mo | S3 only — you still run GeoServer | ~$2,760/yr (storage) |
| smallest.zip Geospatial | $0.005/GB-mo on compressed bytes | 10 TB compresses to ~3–5 TB typical | ~$300/yr |
10 TB raw × ~40% average ratio = 4 TB stored × $0.005/GB-mo × 12 = ~$240. Plus $0.01/GB × 10 TB ingest one-time = ~$100. Numbers assume mixed RGB GeoTIFF + airborne LAS + shapefile / GeoJSON.
That's ~9× cheaper than S3-only storage, ~40× cheaper than Esri ArcGIS Online, ~60× cheaper than Pix4D Cloud — before you factor in GeoServer hosting.
Four formats, one pipeline
GeoTIFF (single + multi-band)
Ortho mosaics, DEMs, satellite, multispectral. Band-separated predictive coding. Pixel-exact lossless for integer and float bands. COG-style tile output supported.
LiDAR (LAS / LAZ)
Airborne and terrestrial point clouds. Scan-order encoding, per-tile spatial index for bbox queries. 6.99% of raw on real airborne — beats LAZ.
Shapefile (.shp + .dbf + .shx)
Polygon, line, point shapefiles with all DBF attributes. TopoJSON-style shared-arc topology. ~9.8% on Natural Earth 10m.
GeoJSON
Feature collections of any geometry type. Bbox-quantized integer grid with shared-arc topology. ~9.85% — beats TopoJSON+gzip (12.2%).
The honest lossy / lossless story
Per-format. No marketing weasel.
| Format | Default | What's lossy | Lossless mode |
|---|---|---|---|
| GeoTIFF integer | lossless | nothing — pixels byte-equal | default |
| GeoTIFF float (DEM) | lossless | nothing — IEEE 754 bit-exact | default |
| LAS / LAZ | lossless at LAS scale | nothing below the source file's declared XYZ scale (e.g. 0.01 m) | default |
| GeoJSON | quantized to 1e6 grid (~5 m on continental bbox, ~5 cm on city bbox) | coords snap to integer grid; collinear vertices removed via Douglas-Peucker | quantize=0, precision=8, simplify=0 — sub-cm grid, no DP |
| Shapefile | quantized to 1e6 grid | same as GeoJSON; DBF attributes preserved exactly | quantize=1e8 for ~5 cm grid |
Benchmarks on real geospatial files
All results from our public validation harness — see codec-audit/geospatial.
| File | Original | Compressed | Reduction | Roundtrip |
|---|---|---|---|---|
| Synthetic 4-band 512×512 uint16 GeoTIFF | 2.00 MB | 9.9 KB | −99.5% | pixel-exact |
| Planet Labs UDM (3-band uint8) | 8.4 KB | 360 B | −95.8% | structural |
| Vera Rubin J1339 (3-band RGB) | 4.14 MB | 2.55 MB | −38.5% | pixel-exact |
| Natural Earth 10m countries (GeoJSON) | 24.1 MB | 2.37 MB | −90.2% | geometry @ 5 m grid |
| Synthetic 10K polygons (GeoJSON) | 4.45 MB | 367 KB | −91.9% | structural OK |
| Natural Earth 10m countries (Shapefile) | 11.6 MB | 1.14 MB | −90.2% | all attrs preserved |
| Synthetic 2K point shapefile | 346 KB | 26.9 KB | −92.2% | 2000/2000 recs + shapes |
| Airborne LAS (736K points) | 14.7 MB | 1.03 MB | −93.0% | lossless @ 0.01 m |
| Synthetic terrain LAS (200K) | 6.49 MB | 2.06 MB | −68.2% | 200000/200000 pts |
Vera Rubin imagery is photographic RGB — high-entropy — and is closer to a worst case. Drone ortho mosaics typically sit between Vera Rubin and the synthetic raster.
See it on your own data
Drop a .tif, .las, .laz, .geojson or zipped .shp. Watch it shrink. Verify the roundtrip in QGIS, GDAL or laspy yourself.
No signup. No tracking. Up to 200 MB / file.
Frequently asked questions
Is the CRS preserved?
Yes. We store the WKT / EPSG of every file in our binary header and re-emit it on decompress. GeoTIFF GeoKeys, .prj sidecars and LAS coordinate system VLRs all round-trip.
What about TIFF / GeoTIFF metadata?
Tags 256-339 (image structure), 33550 / 33922 (model tiepoint & scale), 34735-34737 (GeoKey directory), and standard GDAL metadata XML are preserved. Some vendor-private tags are passed through verbatim; obscure proprietary tags may be dropped — file an issue and we'll add them.
Is it ESRI-compatible?
Yes. Output is a stock .shp / .shx / .dbf bundle that ArcGIS, QGIS, MapInfo and GeoPandas all open natively. We do not invent a proprietary format that locks you in — the compressed .shz decompresses back to a vanilla shapefile.
Does this integrate with GDAL?
The decompressed files are standard GeoTIFF / LAS / SHP / GeoJSON — GDAL reads them with no driver changes. For inline use, our Python SDK exposes open_gtz() that yields a numpy array or a rasterio-style band reader.
What about Cloud-Optimized GeoTIFF (COG)?
We support COG-style output: tiled internal layout, internal overviews preserved, IFD ordering compatible with the COG spec. Range-request access is on the roadmap — today you fetch the whole compressed file (which is much smaller than the COG it came from anyway).
Drone software integration?
Works downstream of DroneDeploy, Pix4D, OpenDroneMap and Bentley iTwin — whatever they export (GeoTIFF, LAS, SHP, GeoJSON) we compress. No plugin needed for the export step; for archival pipelines we provide a CLI and a Python SDK.
Bbox queries on big LAS files?
Yes — our LAS format stores per-tile bounding boxes in the header. A center-region query decompresses only the overlapping tiles, not the whole file. 4×4 grid by default; tune with --tile-grid.
On-prem / air-gapped?
Available on enterprise. Single binary, no callbacks, no telemetry. Bring your own object store (S3, MinIO, Ceph, Azure Blob, on-prem NFS).
What if smallest.zip disappears?
The decompressor is a standalone open-source binary — we publish it and the file-format spec under a permissive license. Your data is never trapped.
How does it compare to LAZ / Cloud-Optimized GeoTIFF?
On real airborne LAS we get 6.99 % of raw vs LAZ's 8.2 % — ~15 % smaller. On COG, the internal pixel data uses the same DEFLATE / ZSTD codecs we beat by 30-50 % via predictive coding.
SLA?
99.9% uptime on the standard tier; 99.99% with multi-region replication on enterprise.
Pricing?
$0.01 / GB processed + $0.005 / GB-month stored. First 5 GB each free. See pricing for tiers.
Stop paying S3 + Pix4D + Esri rates for cold geospatial data
Drop a real GeoTIFF, LAS or shapefile and see the compression on your own data. No signup, no credit card.